There doesn’t seem to be a lot of activity on these forums recently, but hopefully someone will see this.
I have been using Rhasspy 2.5 for several years to handle the speech components of an open source project. I know that there is some preliminary work on Rhasspy 3, and it makes sense to not spend time adding features or fixing things for 2.5 (for example it seems that Raspberry Pi 5 support has some issues). On the other hand, even if I could figure it out myself, it seems 3 might be in too early of a state for the other project members to switch.
I do eventually intend to add support for Home Assistant, as the open source project is very similar to a smart device. However, it’s in the middle of a pretty big rewrite at the moment and I was sort of thinking of having a standalone version that was independent of Home Assistant as well.
So my question is, would I be better off moving away from Rhasspy and towards Home Assistant (maybe completely)? Is it possible to run Home Assistant on the same Raspberry Pi for the speech components similar to the way Rhasspy works without having to get fully into the Home Assistant development until I’m ready to do that?
I know that I could probably ask some of these questions of the Home Assistant community, but I believe there are people you’re familiar with both.
I was in a similar situation as you about 6 months ago. I have used rhasspy for probably 7 years. One day I ran into an issue and I was looking into solutions and I saw so many things about switching to home assistants voice assistant. So I finally gave in and decided to try it. I will say I am happy that I switched. It’s not that home assistant voice was way better. It was just nice to consolidate.
I have been using home assistant for the last couple years so I already had a home assistant server set up and running. But I had nothing set up for the voice assistant part. So I started digging into setting that up and it was not hard at all.
To answer some of your question, no you dont have to go all in on home assistant. You should look up wyoming protocol. I use the docker version of home assistant, the speech to text, text to speech and wake word components that my home assistant voice assistant uses are all separate docker containers as well. So you can set up those containers and use them completely independent of home assistant. The wyoming protocol also has satellites. None of my satellites have anything home assistant installed on them.
Here is the github for the wyoming protocol:
It almost seems branded for home assistant but you can definitely run each component of it separately and integrate it into something else
I was sort of sad when I shutdown my rhasspy instance. I also thought having things separate were nice. Like if home assistant went down, rhasspy would still be up and I could atleast still use my speakers. I traded separation for redundancy and setup backup instances of home assistant so if my server went down, I could bring up a different one
So long story short, you can’t go wrong either way but by looking into wyoming protocol will probably give you a nice option that maybe is in-between
Thank you so much for sharing this. As I said, I’m in a really awkward spot with my project at the moment, but I’m going to try to digest what you have said and I may have follow-up questions later.
All of my efforts these days are going towards the Wyoming protocol with a focus on Home Assistant. Something that’s missing for non-HA users is a Wyoming server that runs a pipeline with different Wyoming clients (wake word → STT → intent → TTS). It would also need to communicate with Wyoming satellites, which is a bit more complicated but not too difficult.
As you can tell, I’ve spread myself too thin to keep up with everything
I think it’s time to find one or more community members to maintain Rhasspy while I focus on Wyoming.
@synesthesiam No worries. I completely understand your priorities.
In my case, Home Assistant is actually a very good fit, as the project is intended to be a smart home device. I’m just not in a position to switch at the moment. That said, I’m sure there are definitely users who would benefit from Rhasspy updates, if maintainers can be found.
